MiniCPM5-2B
公开成绩与计算依据
综合参考分根据已公开的评测资料整理。查看发布日期依据
综合参考分 51.9 · 16项基准
推理求解 47.1
AIME · 86.5
huggingface.co · 官方发布比较 · 2026
MiniCPM5-2B · Official published comparison; AIME 2026; exact column MiniCPM5-2B. Footnotes and all competing runs retained in source.
MATH-500 · 94.6
huggingface.co · 官方发布比较 · 500-item subset
MiniCPM5-2B · Official published comparison; MATH-500; exact column MiniCPM5-2B. Footnotes and all competing runs retained in source.
GPQA Diamond · 0.702020202020202 比例
Artificial Analysis · 独立评测 · Diamond
MiniCPM5-2B · Current AA standalone protocol. Status verified in aa-method-current.md section gpqa-diamond
HLE · 0.038 比例
Artificial Analysis · 独立评测 · Diamond
MiniCPM5-2B · Artificial Analysis current published protocol; hle-diamond
工具与任务执行 34.0
tau3-Banking · 0.208247422680412 比例
Artificial Analysis · 独立评测 · 1.0.1
MiniCPM5-2B · Current AA standalone protocol. Status verified in aa-method-current.md section tau3-banking
GDPval-AA · 714.95 分
Artificial Analysis · 独立评测 · 2.1
MiniCPM5-2B · Artificial Analysis current published protocol; gdpval-aa-normalized
Terminal-Bench · 0 比例
Artificial Analysis · 独立评测 · 4.0
MiniCPM5-2B · Artificial Analysis current published protocol; terminalbench-4-0
AutomationBench · 0.018006007862431707 比例
Artificial Analysis · 独立评测 · current published task suite
MiniCPM5-2B · Artificial Analysis current published protocol; automationbench-aa
AA-Briefcase · 423.29 分
Artificial Analysis · 独立评测 · 1.1
MiniCPM5-2B · Artificial Analysis current published protocol; aa-briefcase
BFCL · 66.6
MiniCPM5-2B · Official published comparison; BFCL v4; exact column MiniCPM5-2B. Footnotes and all competing runs retained in source.
编程 100.0
LiveCodeBench · 69.1
huggingface.co · 官方发布比较 · release_v6
MiniCPM5-2B · Official published comparison; LiveCodeBench v6; exact column MiniCPM5-2B. Footnotes and all competing runs retained in source.
语言与信息处理 32.1
IFBench · 66.3
huggingface.co · 官方发布比较 · current published task suite
MiniCPM5-2B · Official published comparison; IFBench; exact column MiniCPM5-2B. Footnotes and all competing runs retained in source.
Multi-IF · 71.8
huggingface.co · 官方发布比较 · current maintainer dataset
MiniCPM5-2B · Official published comparison; Multi-IF; exact column MiniCPM5-2B. Footnotes and all competing runs retained in source.
GDP.pdf · 0.014 比例
Artificial Analysis · 独立评测 · current published task suite
MiniCPM5-2B · Artificial Analysis current published protocol; gdp-pdf
IFEval · 86.7
huggingface.co · 官方发布比较 · current maintainer dataset
MiniCPM5-2B · Official published comparison; IFEval; exact column MiniCPM5-2B. Footnotes and all competing runs retained in source.
MMLU-Pro · 70.8
huggingface.co · 官方发布比较 · current maintainer dataset
MiniCPM5-2B · Official published comparison; MMLU-Pro; exact column MiniCPM5-2B. Footnotes and all competing runs retained in source.
资料来源
- 版本日期来源 · 版本资料 · 2026-10-29查阅
- huggingface.co · official_release_comparison · 2026-10-29查阅